<p>Rodrigo is Director of Partner Success at LenddoEFL.</p>


Rodrigo Sanabria

Utilizing alternative data to manage loan portfolios after disbursement 

Blog Post

<em>This is the third and final post in a mini series responding to the CFI report, </em><a href="https://www.centerforfinancialinclusion.org/accelerating-financial-inclusion-with-new-data-2">Accelerating Financial Inclusion with New Data</a>. Read the previous posts <a href="https://content.centerforfinancialinclusion.org/to-bank-the-unbanked-start-using-alternative-data/">here</a> and <a href="https://content.centerforfinancialinclusion.org/aim-build-leverage-partner-persevere-5-tips-to-leverage-alternative-data-to-bank-the-unbanked/">here</a>.

<em>Accelerating </em>describes very interesting alternative data use cases, but I could not help to notice that it is focused on assessing prospects at the point of loan origination. It leaves out how to use alternative data to manage a financial inclusion loan portfolio after the client has received a first loan.

When we think about financial inclusion in a lending context, we naturally tend to think about how to approve prospects with little or no credit history. But when we step back a bit, we soon realize that origination is just a piece of the opportunity. We all know very well that collections and client retention are at the core of a sustainable financial inclusion business.

Financial institutions have used traditional data beyond origination for a long time. Most use bureau data, but many universal banks also use internal transactional data to profile customers. Large banks’ analytics departments and their consultants have become quite skillful at applying behavioral credit scoring and insights to price, cross selling, mitigating churn, and collecting.
<h3>Can MFIs use transaction data for credit scoring and insights as banks do?</h3>
Not so much. On one hand, most MFIs do not have transactional data beyond repayment. On the other hand, they are usually the only one reporting data to the credit bureau about most of their customers. So, credit scoring and insights based on traditional data bring limited benefits.

But MFIs can leverage alternative data to obtain similar benefits. I see two areas of opportunity: Cost effective collections and managing churn, the attrition of customers.
<h3>Leveraging alternative data for cost-effective collections</h3>
New-to-credit customers require more follow up than seasoned borrowers in order to secure payment. Many MFIs’ approach to deal with this challenge is to get loan officers involved in collections. Loan officers usually develop a good feeling for who in their portfolio needs to be reminded of an upcoming installment, and how to prioritize recovery efforts among customers in arrears. But it is an "art" that might be difficult to develop and expensive to operate.

We have worked with a client on ranking customers that are current on their payments by the probability of default in their next payment. We combine their repayment experience (traditional data) with their changes in mobile phone usage (alternative data). This ranking helps our client set differentiated preemptive collections strategies that allow them to focus their resources where they can extract the most value. Portfolio quality has improved while the investment in collections has stayed stable. I believe this will lead to higher approval rates in the future, as well as the ability to invest more in collections, fostering healthy portfolio growth.

We have also worked on segmenting customers in arrears by their probability of recovery. We combine traditional repayment data with alternative data collected at origination in order to separate customers in three categories: (1) those that will most likely get back on track with little or no intervention, (2) those that are likely to recover if loan officers intervene on time, and (3) those that will most likely not be recovered anyway. By focusing on the second group, loan officers improve their portfolio performance significantly.
<h3>Building insights to manage churn</h3>
In the financial inclusion business, small loan amounts and short terms of payment make it very difficult to recover the customer cost of acquisition on the first loan unless we charge prohibitive interest rates. Thus, we need to put special emphasis on extending the customer lifecycle in order to build a financially sustainable portfolio.

[caption id="attachment_32967" align="alignright" width="282"]<img class="wp-image-32967" src="https://content.centerforfinancialinclusion.org/wp-content/uploads/sites/2/2018/10/churn-683x1024.jpeg" alt="" width="282" height="422" /> Churn: good for dairy farmers, bad for financial service providers.[/caption]

Many of my clients complain that once they go through the cost and pain of capturing a new-to-credit customer, they lose the client to another institution offering better terms, which they can do because the customer's credit history has been enriched.

SMS and email data provide great insights into customers’ financial relationships and habits. These data may be used to both develop products that better fit specific segments of an institution's client base, as well as to target clients with the right offerings better suited to their needs and preferences; and to do it at the right time to prevent this sort of churn.

For example, one may obtain an indication of income growth, increased business activities, or seasonal working capital requirements that may trigger a new credit offering. We may also be able to spot clients "shopping around" in order to target them with retention offerings.

In a prior <a href="https://content.centerforfinancialinclusion.org/aim-build-leverage-partner-persevere-5-tips-to-leverage-alternative-data-to-bank-the-unbanked/">post</a>, we talked about the challenge of gathering large datasets in order to build origination models. The good news here is that a good way to accelerate the capture of value from alternative data is to look beyond origination into collections and client retention. Customer insights and collection scores can be developed quickly by combining new alternative data with traditional data that the institution already has.

In summary, the journey to harness alternative data value can be accelerated by focusing first on using it on the current client base in order to foster retention and improve portfolio performance, while at the same time gathering data on new clients to build origination models that allow us to approve more people while protecting performance.

Beyond Origination

Alternative data can help FSPs reduce loan defaults and speed up the approval process, but pitfalls exist.

I have been rolling out alternative data initiatives for financial inclusion across Latin America for several years. At some point, my clients ask: “is this going to work?” My usual answer is “I’ve failed enough times to have figured this out.”

This is a fairly new and not completely mature field. LenddoEFL has been doing this for over 10 years. While there is still a lot to learn, my team and I can share some wisdom.

In response to <a href="https://www.centerforfinancialinclusion.org/publications-a-resources/browse-publications/1030-accelerating-financial-inclusion-with-new-data">Accelerating Financial Inclusion with New Data</a>, I recently <a href="http://cfi-blog.org/2018/08/14/to-bank-the-unbanked-start-using-alternative-data/">wrote</a> about the promise and challenge of using alternative data to bank the unbanked. We’ve learned a lot about applying alternative data and have identified five key success factors:
<ol>
 	<li>Aim at the pain</li>
 	<li>Build on top of your current business</li>
 	<li>Leverage the best data source for you</li>
 	<li>Partner with somebody that can handle multiple data sources</li>
 	<li>Persevere. Capture low-hanging fruit without losing sight of the big prize</li>
</ol>
We will tackle one at a time.
<h2>1. Aim at the pain</h2>
Some financial institutions come to us interested in “trying out” alternative data. Our usual question is “what problem are you trying to solve?” Sometimes they are not clear about what they want to solve, and sometimes they want to fix too many things at the same time. The whole approach for the initiative will depend on this understanding. Choose one pain, focus on it, and build the KPIs to measure success according to this.

Keep repeating to everybody the pain you are attempting to solve to make sure everybody shares the same understanding.

These are some examples from our experience:

• An MFI wanted to increase productivity per loan officer while maintaining default rates: reduce turn-around-time, workload in the field, and complexity. Its client base was made up of unbanked and thin-file customers, so, automation based on traditional scores was not an option. Solution: <a href="https://www.youtube.com/watch?v=c0b57p9lnm8&amp;feature=youtu.be">Collect psychometric information for credit scoring which would allow a centralized, automated process</a>.

• A non-traditional microlender wanted to obtain early warnings of clients that would likely fall in arrears on their next installment so that they could better focus pre-emptive collections efforts. By combining traditional repayment data with Android phone data, we are able to “rank” clients by the probability of next payment default. Now they can focus on the the one-third that will create 75 percent of the defaults.

• A traditional financial institution was turning down about one-third of applicants due to lack of credit history, and not belonging to the “right” demographics. They decided to invite “rejects” to re-apply by providing psychometric information, which allowed us to “rescue” about half of those prospects <a href="https://www.eflglobal.com/casestudies/banco-pichincha/">without increasing the default rate</a>.

• A home appliances retailer providing $200 loans to consumers was losing clients due to the time required to verify their identity. By leveraging social network data, they have been able to reduce the approval turnaround time from two days to a few minutes in most cases. They have been able to approve more clients, reduce the cost of identity verification, and reduce cases of fraud.
<h2>2. Build on top of your current business</h2>
A good friend and a brilliant risk professional called me asking for help: “We are planning to launch a new product, for a new segment, in a new channel, so we need to use a new source of data to build an origination model.”

“Too many ‘news’ in the equation,” I told him. However, I joined his new venture.

You can guess how this adventure ended: slow volume uptake, lack of an actionable model after several months, and little enthusiasm to keep investing in order to capture value.

As we discussed in the <a href="http://cfi-blog.org/2018/08/14/to-bank-the-unbanked-start-using-alternative-data/">first post</a>, building models with alternative data is a numbers game. You need volume.

In the successful cases we mentioned before, we collected alternative data from a population that was already being served through a channel already established. This was to support a product with existing traction in the business. Innovation was concentrated in the data source and methodology to asses risk.
<h2>3. Leverage the best data source for you</h2>
Each source of data has advantages and drawbacks. In the front end, some sources may create more or less friction on the client onboarding, depending on origination processes. On the backend, usually the “low-friction” data is not structured. Unstructured data is not organized in a predefined way, so using it to build a risk model is more challenging than using structured data.

Once you have identified the pain point, you may work out with your partner/vendor the tradeoffs considering your population and channel. Note the following tips:

• Highly digital populations already served through an online channel may be approached using digital data, but you must make sure that you can get the volumes required to build a model based on unstructured data (unstructured data requires more volume to build a model).

• People with whom you already have an ongoing relationship may be a good population to leverage mobile phone data, as they may perceive a benefit to downloading and keeping your mobile application.

• Less digitized populations, served through traditional channels (branches or field loan officers) may be better suited for psychometrics.

Avoid the pitfall of falling in love with a specific data source and then figure out a use case within your business. Go the other way around: “given my business need, what data source better fits it?”
<h2>4. Partner with somebody that can handle multiple data sources</h2>
“When you only have a hammer, all problems look like nails,” my first boss told me a long time ago. To avoid the pitfall described on recommendation three, you must partner up with a vendor that can manage several data sources.

This will not only let you choose the right pain and business to focus on, but also give you flexibility as you roll out.

For example, we found, while working with a one client that their clients would willingly share their email data. Unfortunately, we found that they used their email so scarcely, that we couldn’t score many of them. Now we are working with psychometrics in this population.

In another situation, we started using psychometrics to approve more people at a Mexican e-lender. In the meantime—while they were approving more clients—we collected digital data from these same applicants. After several months, we have been able to combine both sources of data to approve even more people.
<h2>5. Persevere</h2>
If you are like most of us and work for an organization that needs results in a few quarters, structure your initiative to collect early results that may give you inertia while you go for the long-term prize.

We work with an institution that provides big loans. They do not have that much volume, but they invest heavily in each prospect. Big stakes, low volume is the most challenging environment to build an alternative data-based score. It took us almost 4 years, but now they are <a href="https://www.youtube.com/watch?v=XkdViqZrFiE">harvesting the fruits of their perseverance.</a>

To deal with this issue, you need to be creative to identify secondary pain points that may be addressed quickly along the way.

For example, we worked for a retailer that wanted to increase approvals while keeping defaults in line by approving new-to-credit consumers. Loans had mostly 24 to 36-month terms and most 60 days defaulters tended to recover. That was a challenging situation: we would have to wait 12 months for vintages to mature, and look for 90 or 120 days in arrears for the “bads” to profile. It looked like a 2 to 3 year project.

But we found a secondary pain: “straight rollers.” These were loan recipients who didn’t pay their first two or three installments and were eventually written off. We collected data on all their clients to quickly build a “straight rollers model.” We only needed 3 installments on each vintage to identify bads.

Along the way, we are collecting data that will be used to build an admission score to address the main pain.

In summary, building credit policies based on alternative data is challenging. Fortunately, there is enough learning accumulated in our community to avoid some pitfalls and we hope you find these tips useful.

<em>Image Credit: <a href="https://www.flickr.com/photos/ollycoffey/44178312761/in/photolist-2aiTpeB-29yx9Jy-CShyD1-ubkKL-Gwk8J-TZjtaB-chvbey-VkyiSs-UQQQVR-23MtnU2-7UaJJt-4PBwQM-78Hrup-VkraHG-9NFy2H-YrYmC9-zMd4V-CLUsy-4c2Cg1-MFuo7f-hGbHNz-nLob8w-dfsjwh-dfsdYV-S91RPs-25NPKA5-3HvEGy-sP6gPL-ePB39d-q7fEng-2g9AMe-Vj4Ktb-dQv6D9-93gQWB-6PKUAu-nj6bUc-L9Q77G-28xXGRL-7h11tB-FCi2qp-27rQtYj-VuZD4L-U8a7JG-VA2AxC-W6jpzN-QGVNSh-ee9ipB-8UEP3j-21btTU9-24VtQzX">Olly Coffey</a> / <a href="https://creativecommons.org/licenses/by-nc/2.0/">CC by 2.0</a></em>

&nbsp;

Aim. Build. Leverage. Partner. Persevere: 5 Tips to Leverage Alternative Data to Bank the Unbanked

Capturing digital footprints using psychometrics can help FSPs reach the unbanked.

<em>This post is one of five entries related to the report,<a href="https://www.centerforfinancialinclusion.org/publications-a-resources/browse-publications/1030-accelerating-financial-inclusion-with-new-data"> Accelerating Financial Inclusion with New Data</a>, a collaboration between CFI and the <a href="https://www.iif.com/">Institute of International Finance (IIF)</a>.</em>

In a recent <u><a href="https://cfi-blog.org/2018/06/25/accelerating-financial-inclusion-with-new-data/">post</a></u> on her report, <em><a href="https://www.centerforfinancialinclusion.org/publications-a-resources/browse-publications/1030-accelerating-financial-inclusion-with-new-data">Accelerating Financial Inclusion with New Data</a></em>, Tess Johnson highlighted the huge opportunity that alternative data represents for the future of financial services. The simple fact that mobile and internet penetration have surpassed financial services penetration in most emerging markets hints at a big opportunity: many people who have had no meaningful access to formal financial services are creating digital footprints financial service providers can capture and analyze to reach them with commercially viable services that help them improve their lives. This prospect is also made possible thanks to machine learning and big data methods that were not available to us a few years ago.

For those of us in the world of financial inclusion, these are very exciting times: the simultaneous emergence of online penetration and data analysis methods is generating an opportunity that our predecessors in this field couldn’t even have imagined.

The bad news is that harnessing digital footprint data using machine learning is not easy; it requires time, commitment and skills that are in short supply. However, the good news is that those with the vision and  endurance to leverage this opportunity will build a competitive advantage that will be sustainable for years to come.

When developing an alternative credit score based on traditional information (e.g., demographics, repayment data), analysts usually have historical data to design and train models. Through back testing, the credit scoring model is applied to historical data to see how accurately it would have predicted the actual results (i.e., loan repayment). We can get a pretty good sense of how the model will perform in the future and set up a credit policy accordingly. Yet, when we cannot use such traditional data sources, we are entering into uncharted territory.

Lacking prior information about our current customers' psychometric profile or digital footprint, we must build those data sets from scratch. Depending on the data source, we may need very large data sets to compensate for the lack of data structure (unstructured data is simply data that is not easily accessible in a format or structure, like an Excel spreadsheet, that is optimal for generating insights). Just as with all other artificial intelligence applications, the more data you collect, the more predictive and stable your algorithms become. LenddoEFL is an example of an organization that gathers data for these profiles and footprints. It is an alternative credit scoring and verification provider that uses psychometric and other data about a loan applicant to determine a credit score and verify identity.

Furthermore, even state-of-the-art alternative data sources do not necessarily allow you to build models that are stable and reliable across multiple segments of the market. Therefore, you need to build algorithms that are specific to your target population.

One of the most challenging issues when implementing alternative data scoring initiatives is showing the results that can be achieved within a given set of time and budgetary constraints. In the long run, after the portfolio has matured, you can show whether using alternative data allowed you to approve more applicants within your target default levels, controlling by business cycle. But if you are working with 24- to 36-month loans, it may take three or four years before you can fully assess the impact of using alternative data, by which time internal attention spans may have already run short.

To account for that, LenddoEFL uses early indicators of model performance. We set a target maturity and days in arrears according to a financial institution’s portfolio’s profile, for example, 60 days in arrears within the first 9 months. Then we calculate a Gini coefficient—a scale of predictive power that can help lenders understand how good its credit score is at assessing who will repay and who will default on a loan (not to be confused with the Gini coefficient that measures income inequality) for the model as applied to that portfolio. (For more details on how to use the Gini, check out our blog series from our risk and analytics team: <a href="https://include1billion.com/news/2018/6/21/blog-on-the-use-and-misuse-of-gini-coefficients-in-credit-scoring">Part 1</a>, <a href="https://include1billion.com/news/2018/6/26/blog-on-the-use-and-misuse-of-gini-coefficients-in-credit-scoring-comparing-ginis">Part 2</a>, <a href="https://include1billion.com/news/2018/7/10/blog-on-the-use-and-misuse-of-gini-coefficients-in-credit-scoring-gini-and-acceptance-rate">Part 3</a>).

Is it too late to pursue an alternative credit scoring initiative? There are plenty of companies already doing this—<a href="https://youtu.be/c0b57p9lnm8">Te Creemos in Mexico</a>, <a href="https://www.youtube.com/watch?v=RX4tJTLic0Y">Mynt in the Philippines</a> and <a href="https://youtu.be/XkdViqZrFiE">Business Partners in South Africa</a>—but only a few lenders are utilizing alternative data in each market. You could be the first institution in your segment and country to implement such an initiative, and you can still take advantage of others’ experiences and learning.

The sooner you start collecting data and building models, the sooner you will be able to underwrite the unbanked segment better than your competition, and the longer the window of advantage will be. For those who start late, catching up with the early adopters will be a great challenge.

Stay tuned for my next blog post where I will outline recommendations for getting starting using alternative data!

&nbsp;

&nbsp;

&nbsp;

&nbsp;

Rodrigo Sanabria

Latest by Rodrigo Sanabria

Stay informed. Subscribe to our newsletter.